DOA Estimation Of Indoor Source Sounds Based On Beam Design And Spherical Harmonic Domain Analysis

Direction-of-Arrival(DOA)estimation of sound sources is a key technique for indoor multi-source localization and tracking.In indoor environments,sound signals can be collected using a spherical microphone array.And combined with the theory related to the Spherical Fourier Transform(SFT),the sound signal is analyzed and processed in the spherical harmonic domain.This is an effective measure to solve the DOA estimation of indoor multiple sound sources in low signal-to-noise ratio(SNR)and high reverberation(RT).For the low SNR and high RT environment,this paper intends to solve the acoustic localization problem of multiple sound sources,adjacent sound sources and the presence of interference sources.Indoor multi-source DOA estimation research is carried out in terms of three-dimensional beam design of spherical microphone array and spherical harmonic domain spectrum estimation.The main work and contributions of this paper are as follows:1.A Spherical Harmonics Domain-Multiple Signal Classification(SHD-MUSIC)algorithm(abbreviated as SHD-BMUSIC)is proposed to address the problem of poor localization accuracy due to the influence of reverberant noise in multi-source DOA estimation.The method is capable of transforming the spherical harmonic domain acoustic pressure to beam domain output by pre-forming multiple beams and constructing beam domain MUSIC space spectrum for peak search.Taking advantage of the decoupling of the frequency and angular components of the guide vector in the spherical harmonic domain,the smoothing method is used for decorrelation to improve the anti-reverberation performance of the algorithm.In addition,the 3D beam of the spherical array has rotational symmetry and its beam steering is decoupled from the beam shape.It allows arbitrary beam steering without the need to redesign the beam shape,making the 3D beam more flexible.2.The spherical harmonic domain beamforming part of the SHD-BMUSIC algorithm proposed in this paper has significant flexibility.It can be designed with different beam weights to suit different acoustic environments without changing the beam steering.In this paper,beam patterns are designed separately for two multi-source environments with adjacent sources and the presence of interference sources,to make the SHD-BMUSIC algorithm more universal in low SNR and high RT environments.(1)In order to improve the localization capability of SHD-MUSIC algorithm for adjacent sound sources in physical space,we propose SHD-BMUSIC algorithm using multitarget beam pattern.The multi-objective weighted design is represented as a Quadratic Constraint Quadratic Planning(QCQP)optimization problem solved with a trade-off between directionality index(DI)and white noise gain(WNG).Simulation experimental results show that the SHD-BMUSIC algorithm,which is improved using multi-target beam pattern,has better adjacent source discrimination capability compared to the SHD-MUSIC algorithm;(2)In order to solve the influence of directional interfering sound sources on the target source azimuth estimation,the SHD-BMUSIC algorithm is proposed to be improved using the cardioid beam pattern.Depending on the design goals,cardioid beam patterns can be designed to optimize different array characteristics.Such as forming a zero trap across the observation direction,maximizing the directionality index,and maximizing the front-toback ratio.The simulation experimental results show that this SHD-BMUSIC algorithm,which is improved by using cardioid beam pattern,can effectively suppress the interference source and improve the azimuth estimation accuracy of the target source by benefiting from the advantage of better low beam side flap of cardioid beam pattern.